Skip to main content

Understanding Social-Ecological Systems: Coupling population and satellite remotely sensed environmental data to improve the evidence base for sustainable development

Periodic Reporting for period 1 - USES (Understanding Social-Ecological Systems: Coupling population and satellite remotely sensed environmental data to improve the evidence base for sustainable development)

Reporting period: 2015-11-01 to 2017-10-31

Human populations are expanding and changing rapidly in many developing countries. Governments need to keep pace with these changes to ensure that resources are planned and delivered appropriately. For example, ensuring that there are enough school places for children. The traditional way that governments plan resource allocations is through the data collected in a national Census. These are huge undertakings as they cover every person within a country and therefore are time consuming and expensive to plan and undertake. This expense often means that census data collection is conducted just once every 10 years in most countries. This means that by the time the results of a new census are published changes in socioeconomic conditions can be seen but often the data do not contain enough information telling us why or how these changes occurred. In developing countries the majority of rural communities rely on natural resources and environmental products for food, fuel, building materials and medicines.Understanding relationships between poverty and the environment have important policy implications for sustainable human development and ecological conservation. Past attempts to study these relationships have been limited by data availability and a resulting focus on small case study examples from single time periods. The USES project examined population-environment relationships using data with unrivaled detail in household level changes in socioeconomic conditions and environmental characteristics. The overall objective of the project was to couple data derived from satellite images with household surveys to explore ways in which we can increase our understanding of population-environment relationships.

The research is important for society because the quality and quantity of data sets such as satellite imagery are becoming increasingly common at cheaper costs. There is increasing evidence that smaller scale changes and characteristics of landscapes that can be detected in satellite imagery could be used to inform policy makers about potential socioeconomic changes in these regions. For example, in some villages, household roof material changes from straw to metal as incomes rise and peoples confidence in their medium term incomes increase. These changes roof material can be detected in fine-grained satellite images. If these changes can be detected automatically from satellite imagery it may be possible to use this information at the government level to indicate a region that is going through a positive socioeconomic change. This would have several benefits to society including reducing the costs of expensive household surveys to monitor changes between census periods (the satellite data would not replace census or household surveys but would supplement the information available to decision makers). The USES project wanted to examine if there was more to be gained by looking at fine-grained data (household level socioeconomic conditions and 2 m spatial resolution satellite imagery which means individual houses and some vehicles can be detected in the imagery).
The project began with a workshop in New York at Columbia University where data were collected and discussions on the project aims and objectives and any potential problems with datasets were highlighted. The rest of the first year was spent building an approach to linking satellite data with household survey data (see attached figure). We found that there was a complex land use structure within the Kenyan study site and we created a new approach to linking remote sensing to household survey data based on this structure. The approach had four different levels representing four key aspects of land use in the region; (1) homestead; (2) agricultural fields; (3) community resources, and; (4) wider regional area. We compared the new approach with a single level that was traditionally used in the past to see if our new method was better.

Results indicated that our new method was better at predicting household poverty in Kenya than the traditional single polygon approach. Results also indicated that household footprint size detected from fine-grained satellite data was the most important variable for predicting household level poverty followed by the amount of bare-agricultural land in an area. These results are being used to inform future large scale grant applications to upscale our approaches to regional and national levels to see if remote sensing can be used to support government decision making for resource allocations. Other results include, an overview of the changes in household socioeconomic conditions during a 10 year period in Kenya and Tanzania and overview of changes in the environmental conditions (derived from satellite imagery) during the same time period.
Significant progress beyond the state of the art was made in several areas of research. This progress was due to the fact that few studies have attempted to link household survey data (with GPS location information) with satellite imagery before. The approach we developed for linking households with environmental data improved results in Kenya. However, it requires much more work than the traditional approaches to linking so when applying this new method to Tanzania and other sites in the future, changes will have to be made that take into account how the landscape is utilised by households in different locations. This has potentially wider impacts because it means that future studies should consider using a more detailed approach to linking socioeconomic data with remotely sensed data that is focused on how populations use landscapes. It could go much deeper than this in the future by considering land tenure rights, access to land etc. As not all people have the same level of access to environmental resources. But we have shown that accounting for the ways in which households utilize different parts of the landscape can improve the relationships between remotely sensed data and poverty.

An additional outcome was the development of a working relationship between the researcher an international non-governmental organisation discussing how remote sensing data could be used to monitor and evaluate international development projects. The results from the USES project cover only small case study sites which are too small to have an impact on government policy at the moment as there would be questions over upscaling to regional and national levels. Working with an NGO has helped us to consider ideas for upscaling the methods in the future. We are currently working with IFAD to develop new projects that re-use IFAD household survey to further test the ways in which remote sensing data can be used to supplement information in surveys and potentially be used in the future for monitoring changes in socioeconomic conditions